The headline productivity numbers that the federal statistical agencies publish each quarter are being misread, with notable consistency, by both the broader media commentary and by several specific policy conversations that depend on them. The misreading produces real consequences for both private decision-making and public policy.

What the numbers measure

What the headline productivity numbers measure is, at its core, output per hour worked across the broader economy. The measurement framework was designed for an industrial economy and has been progressively extended to capture services-sector output with the kinds of approximations that the underlying data permits.

What the numbers struggle to capture

What the numbers struggle to capture is the kind of services-sector output that does not produce easily-measurable transactions. The categories that have been the largest growth areas of the past decade — healthcare, education, professional services with significant intangible-output components — are also the categories where the measurement framework is weakest.

The implication

The implication is that the headline numbers may be either understating or overstating actual economic activity by margins that are large enough to affect policy decisions that depend on them. Whether the bias is positive or negative in any particular quarter is itself the subject of debate among the most rigorous economists who work on these questions.

What this means in practice

For policy purposes, the practical implication is that decisions made on the strength of single-quarter productivity readings are operating on data that is more uncertain than the published numbers suggest. The decisions that hold up best, on past patterns, are the ones that average across multiple quarters and that supplement the headline numbers with more granular sectoral data.

The honest framing

The honest framing is that productivity measurement is hard, that the methodology is doing its best with limited tools, and that the policy and media conversation should engage with the limitations more than it currently does. The conversation will, on past patterns, continue to treat the headline numbers as more solid than they are; that is not a problem the statisticians can solve on their own.